The Tibetic languages and their classification
(Preprint version, Trans-Himalayan Linguistics, De Gruyter, 2013)
Nicolas Tournadre 1
1. Various meanings of “Tibetic”
The term “Tibetic” has been used in the recent past by some authors in different ways to refer to various intermediate levels of classification within Tibeto-Burman (e.g. Matisoff 2003, Beckwith 2006, Lewis 2009, Blench 2011, de Haan 2011, DeLancey 2012, van Driem 2012, Noonan 2012, Chirkova in press inter alia ). 2 Depending on the author, “Tibetic” may be used to refer to a high level of sub-grouping more or less equivalent to “Bodic” or “Himalayish” or to a lower level of sub-grouping such as “Bodish” or an even lower level as an alternative form for “Tibetan languages/dialects”. “Bodic”, also referred to as “Tibeto-Himalayan” (Michailovsky 2011), groups together many Eastern Himalayan languages (such as the Kiranti languages), Western Himalayan languages (such as Kannauri, Almora), the so-called “Tibetan dialects” and “Bodish languages” (such as Tamang, Gurung, Bumthang, Tshangla etc.). “Bodic” is a “heterogenous and impressionistic” group (van Driem 2011: 31). “Bodish”, sometimes referred to as Tibeto-Kannauri, although less problematic is also not well-defined. It groups together the “Tibetan dialects”, Tamangic languages and some other languages such as Tshangla, Bumthang, Kurtoep, Bake 3 etc. The historical comparative methodology has so far not provided common innovations which would delimit clearly the Bodish subgroup. We can only agree with Matisoff’s (1989, 1990) note of caution to megalo- reconstructionists, or van Driem who states that it is safer “to discover the structure of the [Tibeto-Burman] family tree by working up from the firmer group of lower-level subgroups to the higher levels of superordinate subgroups” (2011: 37). Thus the use of “Tibetic” to replace already existing labels referring to subgroups which are not well- defined or to create new intermediate levels of classification without clear scientific criteria is not appropriate.
1 Aix-Marseille University and CNRS (Lacito). I would like to warmly thank Nathan W. Hill, Guillaume Jacques, Hiroyuki Suzuki, Randy LaPolla, Lauren Gawne and two anonymous reviewers, who read an earlier version of this article for their relevant remarks and suggestions. 2 The Sino-Tibetan macrofamily groups together the “Tibeto-Burman” and “Sinitic” languages. I think it is more convenient and appropriate to call this macrofamily Sino-Tibeto-Burman (STB), because it allows us to clearly include the Sinitic languages (whatever their status may be) when discussing general issues about the inner and outer classification of the Tibeto-Burman subgroup. However, in this paper, I will use the traditional taxon ‘Tibeto-Burman’ (TB). 3 Bake or brag-skad is a language spoken near the Basum lake (brag-gsum mtsho) in the Kongpo area, less than 500 kms from Lhasa.
Because of its various meanings, the term “Tibetic” could become useless, unless it receives a precise definition and is used in a consistent way.
2. Definition of “Tibetic”
The term “Tibetic” could, however, become a useful replacement for the notion of “Tibetan dialects”, which is not appropriate for various reasons. First, the notion of “Tibetan dialects” implies the existence of a single “Tibetan language”. However, the so-called “Tibetan dialects” refer in fact to various languages which do not allow mutual intelligibility. Until the last decades of the 20th century, many scholars would still classify the “Tibetan dialects” into three major groups: Khams 4 , Amdo, Ü-Tsang (thus taking into account only the “dialects” found in China) or into five major dialects: Central (Ü-Tsang), Southeastern (Khams), Western, Southern and Northeastern (Amdo) 5. In recent years, some authors such as Sun (2003a, 2006, 2007), Suzuki (2009b, 2009c, 2009d, 2011a, 2011b), Zeisler (2011, 2012), Haller (2000, 2004, 2009), and Chirkova (2005, 2008b), to name a few, have provided new data on various languages and dialects of the area. These very significant contributions have resulted in new developments in the understanding of the Tibetan linguistic area and its diversity. Second, these “Tibetan dialects” are spoken not only by Tibetans per se but also by other ethnic groups such as Ladakhi, Balti, Lahuli, Sherpa, Bhutanese, Sikkimese Lhopo, etc. who do not consider themselves to be Tibetans 6. They do not call their language “Tibetan”. In a similar way, we do not talk of Latin Languages but of Romance languages and do not think of French, Portuguese, Italian, Catalan or Romanian as various dialects of Latin. 7 With the recent descriptions of many new “dialects” or “languages”, scholars of Tibetan linguistics have come to realize the incredible diversity of this linguistic area. The representation of a single language is no longer viable and we have to speak of a language family. In fact, the Tibetic linguistic family is comparable in size and diversity to the Romance or Germanic families.
4 I use here the orthograph Khams (which corresponds to the Wylie transliteration and not the pronunciation) because the word Kham also refers to a Tibeto-Burman language spoken in Nepal and not related to Tibetan. 5 See Gesang Jumian [1964] 2002; Qu Aitang 1996. For a critical approach, see Sun 2003a or Tournadre 2005. 6 In many areas outside Tibet, such as the southern Himalayas, people speaking Tibetan-related languages sometimes call themselves ‘Bhutia’ ( ་ཊི bhoTi ) or ‘Lamas’ ( ་མ་ bla-ma) . 7 Nor would we think of English as a dialect of German just because German and English belong to the Germanic family. However the term “dialect” is sometimes used in a broader sense and means related varieties: so, for example, one of the definitions of “dialect” given in the Merriam-Webster Online Dictionary (http://www.merriam-webster.com/) is “one of two or more cognate languages : French and Italian are Romance dialects ”.
The term “Tibetic” is thus very convenient to denote a well-defined family of languages derived from Old Tibetan 8. The language called Old Tibetan was spoken at the time of the Tibetan empire (7th--9th centuries). According to Hill “Old Tibetan was originally spoken in the Yarlung valley, the cradle of the Tibetan empire” (2010: 111). While I tend to agree with this statement, it is hard to delimit the precise extension of the area where this language was spoken as well the degree of dialectal diversity at the time of the Empire. Much more controversial is Hill’s remark about the relationship between “Common Tibetan”, the ancestor of all the modern languages (see Section 4 below on Proto-Tibetic) and Old Tibetan: “Because Tibetan languages began to diverge from each other some centuries after Old Tibetan was committed to writing, the written system represents an état de langue (Old Tibetan) older than that reconstructible via the comparative method (Common Tibetan)” (2010: 112). Old Tibetan is very similar to the Classical Literary language which has preserved a very archaic orthography. And indeed, all the modern languages not only have regular reflexes of Classical Literary Tibetan (CLT), they also share a core vocabulary and grammar. The Tibetic language family belongs to a very small circle of language families throughout the world, derived from a common language which is identical or closely related to an old literary language 9. This small group includes for example the Romance languages with Latin, the Arabic languages (or “dialects”) with Classical Arabic, the Sinitic languages with Middle Chinese 10 , the modern Indic languages with Vedic Sanskrit 11 etc. There are only a few other examples of this specific relation between a proto-language and an old literary language. These language groups or families have a great significance not only for the theory of language evolution but also for the typology of sound changes, for morphology and syntax. It is particularly true for Tibetan since some of the attested sound changes have no equivalent in European languages. As Jacques (p.c.) rightly points out, “a general model about the evolution of languages has to take into consideration the specific features of the Tibetic languages. Their interest goes well beyond the Sino-Tibetan macrofamily”.
8 Although in some rare cases such as Baima or Khalong, a Qiangic substratum is a very probable hypothesis. 9 The ideas elaborated in this paragraph were suggested by Guillaume Jacques, whom I thank. 10 Except maybe the Min dialects which bear more archaic features. 11 A few languages have old literary relatives but these literary forms can not be considered as (nearly) identical to the proto-languages: this is for example the case of the Runic inscriptions in relation to Proto- Germanic, Avestan and Old Persian in relation to Proto-Iranian, or Old Church Slavonic in relation to Proto- Slavic. In the case of Iranian or Slavonic languages, the Old Literary languages are precious to reconstruct the Proto-languages but their reconstruction is achieved by the comparative method.
3. Identification of Tibetic languages
The identification of a Tibetic language should thus be based on a number of phonological, morphosyntactic and lexical criteria, which we will mention below. It is important to use such criteria also for the so-called “Bodish” languages, as well as the rGyalrongic and Qiangic languages languages spoken in eastern Tibet (China). These Tibeto-Burman subgroups have been heavily influenced by Classical Literary Tibetan and various spoken forms of Tibetan because of political, cultural and liturgical factors, and together with the southern Himalayan Bodish languages of Nepal, India and Bhutan they can be considered to belong to the “Tibetosphere” 12 , i.e. the sphere of influence of the Tibetan language, culture and religion. In the cases of some languages spoken in the Tibetosphere (such as Bake, Tshona Mönpa, Kheng, Kurtöp etc.), the existence of numerous loanwords from Tibetan may at first give the wrong impression that they are Tibetic languages. While all modern Tibetic developments can be directly derived from Classical Literary Tibetan (or forms very closely related to it) or indirectly derived from it (in the case of local innovations), this is not the case for Bodic or even Bodish developments. The examination of the phonological, lexical and morphosyntactic criteria allows us to sort out Tibetospheric languages from the actual Tibetic languages. These criteria reveal that, beyond the first impression, there is a real gap between the Tibetic languages and other closely related Bodish languages.
3.1. Phonology
Tibetic languages exhibit a great variation on the phonological level. For example, some languages have suprasegmental distinctive features (pitch tone, phonation registers such as creaky voice, breathy voice) while other languages, which have preserved some consonant clusters (or traces of these clusters), do not have such distinctions. However, unlike Bodish languages, all the modern Tibetic languages have regular reflexes of Classical Literary Tibetan. This phonological criterion for the identification of a “Tibetic language” specifies that it is not enough for a given language to have a great number of cognates with Classical Literary Tibetan, these cognates should match regular phonological reflexes of the Literary forms. The regularity of sound changes is of course a
12 I coined this term by analogy with the terms “Sinosphere” and “Indosphere”, proposed by Matisoff (1990). The people speaking rGyalrongic languages are all “Ethnic Tibetans” in that they identify with Tibetan ethnicity. A majority of speakers of Qiangic languages are considered as “Tibetans” and are officially recognized as such by the Chinese state. Other groups of Qiangic speakers are classified by the Chinese government as members of the Qiang Nationality and the Pumi Nationality. fundamental rule of classical historical and comparative linguistics and applies very well to the Tibetic family. Let us exemplify the above statement with two examples. The reflexes of the initial consonant cluster LT found in Classical Literary Tibetan may be quite different in the various modern Tibetic languages, but they are absolutely regular. For example, we find the following reflexes: /lta/, /rta/, /hta/, /t ā/, /lh ā/ etc. as shown below 13 :
LTA ‘look at’ > / lt a/ (Ladakhi, Balti), / rt a/ (“archaic Amdo”) or / ht a/ (“innovative Amdo” 14 ), / tā/ (Ü,Tsang, Khams), / lh ā/ (Sherpa).
LTOGS ‘be hungry’ > / lt oks/ (Balti), / rt ox/ (archaic Amdo) or / ht ox/ (innovative Amdo), / tōʔ/ (Ü,Tsang, Khams), / lh ōʔ/ (Sherpa)
Thus in a given Tibetic language, any word derived from a Classical Literary Tibetan word that contains an initial cluster LT, should have the same derivation. In the above list of LT reflexes, all the modern Tibetic forms contain a /t/, except Sherpa which has a surprising form /lh/. In order to check the linguistic affiliation of Sherpa and confirm that it is a Tibetic language, we should find that all the Sherpa words derived from Literary forms that contain a LT sequence should yield /lh/ in this language. And this is actually the case. The cluster LT in words such as lta ‘look at’, ltogs ‘be hungry’ (see examples above) or lte-ba ‘navel’, ltas-mo ‘show’ etc. all yield the expected reflex /lh/. The regularity of sound changes in Tibetic may again be exemplified by the reflexes of the initial consonant cluster SR : / ʂ/, /str/, / ʈ/, /s/.
SROG ‘life’ > / ʂʂʂox/ (Amdo), / ʂok/ (Ladakhi), / str oq/ (Balti), / ʂōʔ/ (Tsang, Sherpa), / ʈōʔ/ or / sōʔ/ (Ü), / sōʔ/ (Khams)
SRUNG ‘guard’ (verb) > / ʂʂʂoŋ/ (Amdo), / ʂʂʂuŋ/ (Ladakhi), / str uŋ / (Balti), / ʂʂʂūʂ ŋ / (Tsang, Sherpa), / ʈʈʈūŋ / (Ü), / sūŋ / (Khams)
As shown in the two examples above, the retroflex sound / ʈ/ is the normal reflex in Central Tibetan (Ü) and will normally appear in any word that contains a SR initial cluster
13 These examples provide the main reflexes for LT and SR in a sample of Tibetic languages. However, the reflexes given for one language are not exhaustive. They also depend on the dialectal variety. For example, the main reflex of SR in Amdo is /ʂ/ but in Golok Amdo, the reflex is /ʈ/. The same is true for other dialect groups, particularly for Khams. 14 See Nishi 1986. The distinction between Archaic and Innovative Amdo could also be referred to as Pastoralists’ Amdo versus Cultivators’ Amdo. such as sran-ma ‘pea’, srang-lam ‘street’, srab ‘horse bit’, srab-mo ‘thin’, sreg ‘burn’, etc. There are some exceptions in the regularity but they may be explained as loanwords from other dialects or as prestigious pronunciations influenced by the Classical Literary language. For example, the word srog ‘life’ may be pronounced as /s ōʔ/, which reflects the reading pronunciation. In some Tibetic languages, reflexes can be quite remote from the original but they remain regular. For example, the word KLAD-PA ‘brain’ becomes /xlatpa/ (Balti); /l ǟpa/ (Ü); /l ēta/ (Sherpa) and even /l ēp/ (Dzongkha) 15 . In Baima (Zhang Jichuan 1997; Chirkova 2008a) and in many Southern Khams dialects (Suzuki 2009a), the phonetic evolution is even more spectacular. For example, in Baima LAG-PA ‘hand’ becomes /i ɑ/; LUG ‘sheep’ > /y/. However, these changes are totally regular. The initial L yields a glide in Baima and the second syllable is reduced in many contexts (e.g. MCHIL-MA > /d ʒuɛ / ‘spittle’). Tibetic languages have also developed a number of phonotactic restrictions which are not present in neighbouring Bodish languages (such as Khengkha, Tamang, etc.). This is the case for example with the onsets /ml/, /pl/ and / ŋr/, which are not allowed in Tibetic languages, but are found in non-Tibetic Bodish languages.
3.2. Lexicon
In order to define “Tibetic”, a simple test based on the word for ‘seven’ was proposed by Nishi (1986: 849), Beyer (1992: 7), and by Michailovky and Mazaudon (1994: 2). In the Tibetic languages, the word corresponding to ‘seven’ is regularly derived from Classical Tibetan bdun while it is not the case in other languages, even when in the case of Tibetospheric languages closely related to the Tibetic languages. The above test is generally valid and could be used as a first indication, but numbers can always be borrowed. Indeed in Japhug (a rGyalrongic language) the ordinal number 16 has been borrowed from Tibetan: βdɯnpa “seventh” < bdun-pa (see Jacques 2004b). Sometimes the lexical items have distinct etymons but they are all cognates with words attested in Classical Literary Tibetan, whose meanings are equivalent or similar. For example, the verb ‘fear’ in the various Tibetic languages is derived from at least 5 different etyma:
15 The latter example is the result of syllable merger. 16 While the cardinal numeral ‘seven’ has remained a rGyalrongic cognate.
‘fear’ : ɕeʔ (Ü) < ZHED ʈheʔ (Ts) < BRED , ʈāʔ, s āʔ (Kh), ʈāk (Ho), ɕʨ ax (Am) < SKRAG ; dʑiʔ (Tö), ʑiwa laŋ (Sh); dʑiks (La) < ‘jigs , ʑiks (Ba) < ‘JIGS ; ɖo: (Dz) < ‘DROG
Tournadre (2005) provides a list of one hundred lexical items found across the Tibetic languages. The Comparative Dictionary of Tibetan Dialects (Bielmeier et al. in preparation) contains hundreds of roots found across the various Tibetic languages. 17
3.3. Morphosyntax
All the Tibetic languages share a number of fundamental morphosyntactic characteristics, some of which sharply differ from the neighbouring closely related languages. The “pronominalized languages” include many Tibeto-Burman languages belonging to various subgroups such as Qiangic, rGyalrongic and Bodic (Kiranti, Kanauri, Dolakha Newar), but no modern Tibetic language has preserved a system of verb agreement. 18 The ancient verb tense-aspect morphology inherited from Tibeto-Burman has been replaced in the Tibetic languages by a system of auxiliary verbs used with nominalized forms of the verb. Most (if not all) the Tibetic languages have developed a rich system of evidential and epistemic markers, which appear as verb suffixes. Classical Tibetan has a system of 10 nominal cases (see Tournadre 2010). Most of the modern Tibetic languages have preserved to some extent a nominal case system inherited from Old Tibetan, although most of the modern case systems are quite reduced and may use only four cases (ergative, absolutive, genitive and dative). Classifiers are not found in Literary Tibetan and none of the modern languages have developed a system of classifiers, although a few rare classifiers do exist in a marginal way.
17 See also “Etymological notes about widespread Tibetic roots” in Tournadre and Suzuki (in preparation). 18 The question of the existence of agreement markers in Proto-Sino-Tibetan is a very controversial issue. See e.g. LaPolla (1992, 1994) and Jacques who remarks: “Some scholars such as Bauman (1975), van Driem (1993), or DeLancey (2010) argue that such a system must be reconstructed for Proto-Sino-Tibetan, while others such as LaPolla (2003) propose that the agreement systems found in various Sino-Tibetan languages are independent innovations” (2012: 83). There is a general consensus that agreement systems are absent in Literary Tibetan and in the Tibetic languages. Jacques (2010) has mentioned what he believes is one single trace of agreement in Literary Tibetan, but the arguments which are based on the slim evidence of a single verb form zos ‘eat (past)’ are not convincing. Even if it were a trace of verb agreement, it already was reinterpreted in Old Tibetan as a tense phenomenon and did not function as a person agreement.
The grammatical morphemes that are cognate in the various Tibetic languages include the following: a) the negative ma- or m(y)i-19 , b) the nominalizing suffix -pa , c) the auxiliary verbs yin and yod , and d) the genitive and ergative cases -gi and -gis .
4. Proto-Tibetic
It is necessary to distinguish clearly between Pre-Tibetic and Proto-Tibetic 20 (PT). The former is the ancestor not only of Tibetic languages but also of languages belonging to a higher grouping that may correspond partly to the Bodish branch (but possibly also higher groupings of Tibeto-Burman). The latter refers to the direct ancestor of the Tibetic languages. Many authors have used Proto-Tibetan to refer to forms that are indeed Pre- Tibetic. Proto-Tibetic reconstructed forms are often similar or identical to the orthography of Classical Literary Tibetan. The reality is of course somewhat more complex. As Sprigg points out “Shafer would have said that we already know what Proto-Tibetan looks like: it is embalmed in the orthographic forms of Written Tibetan”, but he adds that “none of the dictionaries gives a reliable picture of the phonological structure of Written Tibetan during a given Etat de langue ” (1972: 556). Furthermore, as Hill (2011) has remarked, the phonology of Old Tibetan is not well researched. The main phonological features, which characterize Proto-Tibetic are:
1. The preservation of the prefixes inherited from Proto-Tibeto-Burman. 2. The palatalization of dental and alveolar before y. 3. The change from lateral to dental after m. 4. The emergence of distinctive aspirated initial plosives.
4.1. The preservation of prefixes
The numerous morphological prefixes of Tibeto-Burman, are still clearly pronounced, most probably with an epenthetic vowel in Proto-Tibetic, (see Matisoff , 2003: 97). The main prefixes found in Proto-Tibetic are: *s( ǝ),*d( ǝ)/g( ǝ),*m( ǝ) and *b( ǝ). The numbers gives a good illustration of this phenomenon: *g(ǝ)-tɕik ‘one’, *g(ǝ)-nyis ‘two’, *g(ǝ)-
19 For example, Bake ( brag-gsum skad ) has a negation in a- which is an evidence that it can not be a Tibetic language. As we have seen above, the negation is always derived from ma- or myi- in the Tibetic languages. There is much other phonological evidence such as the lack of palatalization of the dental *ti or the fricative *si or lexical evidences such as the words for ‘meat’ and ‘red’ which are not derived from the pandialectal roots respectively sha and dmar . Despite these anomalies, Bake was however classified as a “Tibetan dialect” by some authors such as Qu Aitang (1996). 20 In some works, the term “Proto-Tibetan” is used. It corresponds to the term “Proto-Tibetic” used in the present paper. su - ‘ three’,* b(ǝ)-ʑi ‘ four’, * l(ǝ)-ŋa ‘five’, * d(ǝ)-ruk ‘six’, * b(ǝ)-dun ‘seven’,* b(ǝ)- rgyat ‘eight’,* d(ǝ)-gu ‘nine’, * b(ǝ)-tɕu ‘ ten’. The prefix ‘s’ used for animals and body parts: * s(ǝ)-dik -pa ‘scorpion’, *s(ǝ)-bal ‘frog’, *s(ǝ)-tak ‘tiger’, * s(ǝ)-b-rul ‘snake’ 21 , *s(ǝ)-pra ‘monkey’, * s(ǝ)-kra ‘hair’, * s(ǝ)-nyiŋ ‘heart’, s(ǝ)-na ‘nose’. Other prefixes such as *d (ǝ)-/*g (ǝ)-, *m (ǝ)-/*r (ǝ)- are also used for body parts as well as some animals: *d(ǝ)-myik ‘eye’, *m(ǝ)-go ‘head’, *r(ǝ)-na ‘ear’. For a discussion about the status of the morphological prefix /d/, see Jacques (2008) and Hill (2011).
4.2. Palatalization of dentals and alveolars before y
Palatalization is one of the main features of Proto-Tibetic. The combinations *ty *ly, *sy were not palatalized in Pre-Tibetic 22 but all these combinations have undergone in Proto-Tibetic a palatalization, which is recorded in the orthography of Literary Tibetan. About the phonemic status of the palatals, see Hill (2011). All the modern languages and dialects have now developed reflexes of these palatalized forms.
* ty > tɕ *g(ǝ)-tyik ‘one’ > PT : * g(ǝ)- tɕ(h)ik > OT : gcig/gchig ག ག་ /ག ག *tye ‘big’ > PT : * tɕ(h )e > OT: che *b(ǝ)-tyu ‘ten’ > PT : b(ǝ)-tɕu > OT: bcu / bchu / *tyi ‘what’ > PT : * tɕ(h)i > OT: ci / chi /
In a lot of Bodic and even Bodish languages closely related to Tibetan, we do not find palatalized forms of * t+y. See for example Bake (Basum lake) /ti/ “what”, /t ɨʔ/ ‘one’ which reflects a stage closer to Pre-Tibetic.
*sy > ɕ
*sya ‘flesh’ > PT : ɕa > CLT : sha *syes ‘know’ > PT : ɕes > CLT : shes *sying ‘wood’ > PT : ɕiŋ > CLT : shing
A lot of Bodish languages such as Tamang, Kurtöp have not undergone this change. tsy > t ɕ *b(ǝ)-tsyat ’to cut’ (past stem) > PT : * b(ǝ)-tɕat > CLT : bcad
21 Sagart and Jacques propose the reconstruction *s-m-rul for some ealier stage (p.c. also compare Hill 2011: 448). 22 See Jacques (2004a) and Gong (1977) for these reconstructions.
*m(ǝ)-tsyil-ma ‘spittle’ > PT : * m(ǝ)-tɕ(h)il-ma > CLT : mchil-ma *m(ǝ)- tsin-pa ‘liver’ > PT : * m(ǝ)-tɕ(h)in-pa, CLT : mchin-pa Among the important innovations of Proto-Tibetic is the palatalization of the lateral /l/ in front of y (see Jacques 2004a). This sound law has been dubbed ‘Benedict’s law’ by Hill (2011: 445).
For example: *b(ǝ)-lyi ‘four’ > PT : * b(ǝ)ʑi > CLT : bzhi *lyi ng ‘field’ > PT : * ʑiŋ > CLT zhing 23 *ldi “flea” > PT : * ldʑi > CLT: lji ‘ji *s( ǝ)-lak(s) “iron” > *l-sak(s) > *l-tsyak(s) > PT : * ltɕaks > CLT : lcags This last mentioned is found in all the Tibetic languages but not in the neighbouring Bodish languages such as Kurtöp / Hla:/ and Bumthap /lak/ (Michailovsky and Mazaudon 1994). The lateral of the sequence *bli is also preserved in many other Bodish or even Sino-Tibeto-Burman languages, such as Kurtöp, Tshona (mtsho-sna), and Old Chinese (see Jacques 2004a).
4.3. The change from lateral to dental after m
The change from * ml to md occurred in proto-Tibetan and its reflex is found in all the 24 modern Tibetic languages. Thus PTB *b/m-la ‘arrow’ > PT : mda > CLT : mda’ Some Bodish languages closely related to Tibetic did not undergo this mutation (see Michailovsky and Mazaudon 1994). Hill calls this sound change ‘Bodman’s law’ (2011: 450 note 12).
4.4. The assimilation of b before nasal n
The change from * bn to * mn is also a characteristic feature of proto-Tibetan.
*bnans > PT : mnans > OT : mnand ‘to suppress’ *bnyan > PT : nyan > mnyand ‘to listen’
This sound law has been discovered by Chang (1971) and Hill (2011: 445-446) has proposed to call it Chang’s sound law.
23 Both ‘ji-ba and lji-ba are found in Classical Literary Tibetan. The Tshig-mdzod chen-mo [Great Tibetan dictionary] gives two slightly different meanings for ‘ji-ba and lji-ba , but they have certainly a common etymology. See also Beyer (1992: 78) and Zhang Jichuan (2009). 24 According to Matisoff’s reconstruction (2003: 50 et passim).
4.5. Emergence of distinctive aspirated initial plosives.
Another characteristic of proto-Tibetan is the emergence of distinctive aspirations for plosives as was shown by Li Fang Kuei (1933) and Jacques (2004b). According to Shafer (1950-1951: 772-773) aspiration in Tibetan [Proto-Tibetic] was originally non-distinctive. Hill calls the sub-phonemic status of aspiration “Shafer’s rule” (2011: 441-442). In Old Tibetan the emergence of aspirated sounds is attested, especially for initial plosives, but it is not yet organized as a system. This was already noticed by Shafer (1950-1951). This can easily be proven by the fluctuation found the orthography of Old Tibetan between aspirated and non aspirated consonants. Ex. : gcig /gchig / ‘one’; phyin-chad/ phyin-cad / ‘from now on’ ci / chi / ‘what’, cu/chu / ‘water’, etc.
4.6. Cases of non-coincidence between Proto-Tibetic and Classical Tibetan
For some words, the classical orthography does not correspond to Proto-Tibetic. For example for the word mig ‘eye’, many dialects of Amdo have a form such as /ɣnj ǝx/ or /mnj ǝx/. Some southern Khams dialects also exhibit an archaic reflex of a sound preceding the labial /m/ (Tournadre and Suzuki, in preparation) . The Proto-Tibetic form should thus be reconstructed *d (ǝ) myik . Fortunately, an archaic orthography dmyig is attested in some old documents. 25 On the basis of many dialects (Tö, Balti, Ladakhi, Sherpa, Gyalsumdo, Lhoke, etc.), we should reconstruct for ‘flower’ PT *mentok, while Classical Tibetan has me-tog. Fortunately, the form men-tog is also attested in Old Literary Tibetan (see Hill 2007: 480 note 8). Reconstructable Proto-Tibetic forms are not always attested, however. For the word sbom-po ‘big (for rope)’, the reconstruction based on some dialects of Tö and Amdo (Ngaba) which have rompo should be PT *s( ǝ)brompo . This form is not attested in Literary Tibetan. In some cases, as shown by Tournadre and Suzuki (in preparation), one could be tempted to reconstruct the proto-Tibetan form, however a phonological innovative rule may provide a better account for the phenomenon. For example, the word khang-pa ‘house’ is sometimes prenasalised in some dialects of Amdo, Khöpokhok, Minyag Khams and Baima (Zhang 1997) and thus we could have proposed to reconstruct PT *nkhangpa . But in this case, the phonological environment may provide a better explanation. 26 A few words with this type of prenasalisation (e.g. tshang ‘nest’, phreng ‘beads’, tshang-ma ‘all’) attested in these dialects originally have an aspirated obstruent
25 This spelling does not appear in Classical Tibetan texts. 26 The phonological hypothesis was proposed by Suzuki (p.c.). initial with -ng final, which was omitted and caused a prenasalisation instead (except Amdo Machu, mGolog). Thus, we have to be careful with the reconstruction of Proto- Tibetic forms when they are based on a single region of the Tibetic area and never rule out the possibility of local areal innovations. Bielmeier et al. (in preparation) raises similar issues about the reconstruction of the common ancestor:
In a number of cases the comparative evidence of the dialects does not lead back directly to the Written Tibetan etymological equivalent. Either the evidence leads to a form previous to the Written Tibetan etymological equivalent, making the Written Tibetan equivalent thus not the “ancestor” but simply an “older relative”, or else we have to accept that certain morphonological or grammatical changes took place within individual dialects. To give an example, we have gyang ‘wall’ in Written Tibetan with regular etymological correspondences in all dialect groups of Tibetan, but in Balti we find rgyang ‘wall’. In such a case we have either to start from a Common Tibetan [PT] *rgyang of which Written Tibetan gyang is a later offshoot, comparable to Purik gyang , or else we may assume that there was an internal Balti development from Common Tibetan [PT] *gyang to Balti rgyang by prefixing r-. We would then have to explain the reason for this prefixing. 27
5. The geographic distribution and status of Tibetic languages
Tibetic languages are spoken above all in five countries: China, Pakistan, India, Nepal and Bhutan. Additionally, Sangdam, a Khams dialect is spoken in the Kachin state of Myanmar. For all the Tibetic languages taken together, the total number of speakers is roughly six million. However, this figure is approximate since there is no precise and reliable census. Here is a list of the Tibetic languages and where they are spoken:
China: Ü-Tsang ( ) , Khams ( ), Hor ( ), Amdo ( ), Kyirong ( ), Zhongu ( ), Khalong ( ), gSerpa ( ), Khöpokhok , Palkyi [Pashi]( ) / Chos-rje ( )28 , Sharkhok ( ), Thewo (
27 Website of the Comparative Dictionary of Tibetan Dialects (CTDT), now accessible at: http://www.himalayanlanguages.org/cdtd 28 Sun (2003b) uses Chos-rje but according to Suzuki (p.c.) dPal-skyid [Pashi] is better suited to refer to a group of four dialects which include Chos-rje .